Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A computing system comprising a demand side platform (DSP) server, the DSP server comprising: one or more processors; at least one memory device coupled with the one or more processors; and a data communications interface operably associated with the one or more processors, wherein the memory device contains a plurality of program instructions including a machine learning model which is executable by the one or more processors and configured to determine an estimate of likelihood of user interaction with a content item, the model having been trained using a set of enriched training feature vectors and corresponding interaction event tags derived from a matched data set generated from records relating to content placement events and records relating to user interaction events retrieved from an online data store, wherein the machine learning model is a logistic regression model comprising a plurality of model coefficients that are stored in a dictionary data structure in which each entry is defined by a key and a coefficient value, and each key comprises a hashed representation of a concatenation of a feature name and a corresponding feature value, wherein the machine learning model implements a regularized logistic regression algorithm that includes ‘follow-the-regularized-leader’ (FTRL)-proximal learning that is trained based on associated hyperparameters to optimize learning accuracy on the set of enriched training feature vectors and the corresponding interaction event tags, wherein each record of the matched data set includes a set of raw feature values derived from an associated content placement event along with a corresponding interaction event tag indicating whether or not an interaction event occurred corresponding with the content placement event, wherein the content placement events and the user interaction events occur within a defined time period, and wherein the plurality of program instructions, when executed by the one or more processors, cause the DSP server to: receive, via the data communications interface, information relating to an online content placement slot; receive, via the data communications interface, information relating to a user to whom content within the online content placement slot will be displayed within the defined time period; compute an enriched estimation feature vector based upon: (i) a content item selected for placement within the online content placement slot that will be displayed within the defined time period, (ii) the information relating to the user, and (iii) the information relating to the online content placement slot; generate, for each feature value of the enriched estimation feature vector, a corresponding key; retrieve from the dictionary data structure, for each generated key, a corresponding coefficient value; determine, by the one or more processors executing the machine learning model, an estimate of likelihood of the user interacting with the selected content item within the defined time period based upon the enriched estimation feature vector and the retrieved corresponding coefficient value, wherein the estimate of likelihood of the user interacting with the selected content item comprises a value representing a probability that the user will interact with the selected content item within the defined time period; and determine, based upon the estimate of likelihood of the user interacting with the selected content item within the defined time period, a bid amount for the selected content item for placement within the online content placement slot that will be displayed within the defined time period based on a magnitude of the value with respect to a bid threshold.
A computing system for predicting user interaction with online content and optimizing bidding in digital advertising. The system includes a demand side platform (DSP) server with processors, memory, and a communication interface. The server hosts a machine learning model, specifically a logistic regression model, trained to estimate the likelihood of user interaction with a content item. The model uses enriched training data derived from matched records of content placement events and corresponding user interaction events within a defined time period. The training data includes feature vectors and interaction event tags, where each record contains raw feature values and an interaction tag indicating whether a user interacted with the content. The logistic regression model employs a regularized algorithm with FTRL-proximal learning, optimized using hyperparameters to enhance accuracy. Model coefficients are stored in a dictionary data structure, where each entry is keyed by a hashed combination of feature names and values. During operation, the system receives information about an online content placement slot, user data, and a selected content item. It computes an enriched estimation feature vector from these inputs, generates keys for each feature value, retrieves corresponding coefficient values, and calculates the interaction probability. Based on this probability, the system determines a bid amount for the content item, adjusting the bid according to the probability's magnitude relative to a predefined threshold. This approach optimizes bidding decisions in real-time digital advertising campaigns.
2. The system of claim 1 wherein the online content placement slot is an ad slot, the information relating to the ad slot and information relating to the user to whom content within the ad slot will be displayed is received along with a bid request message transmitted from an ad exchange server, the content item comprises at least one offer for placement within the ad slot, and the instructions, when executed by the one or more processors, cause the DSP server to: transmit, to the ad exchange server, a bid response message in reply to the bid request message; in response to receiving, from the ad exchange server, a successful bid notification, update an online data store with content placement event data relating to placement of the content item; and in response to receiving a notification of a user interaction with the content item, update the online data store with user interaction event data relating to the user interaction with the content item.
3. A method comprising: at a demand side platform (DSP) server having one or more processors: accessing an online data store to retrieve records relating to content placement events, and records relating to user interaction events, wherein the content placement events and interaction events occur within a defined time period; matching retrieved content placement event records with retrieved interaction event records to generate a matched data set which comprises a plurality of records, each record of the matched data set including a set of raw feature values derived from a content placement event along with an interaction event tag indicating whether or not an interaction event occurred corresponding with the content placement event, wherein the content placement event and the user interaction event occurs within a defined time period; computing, from the set of raw feature values, a corresponding set of enriched training feature vectors; training a machine learning model using the corresponding set of enriched training feature vectors and corresponding interaction event tags, wherein the machine learning model is a logistic regression model comprising a plurality of model coefficients that are stored in a dictionary data structure in which each entry is defined by a key and a coefficient value, and each key comprises a hashed representation of a concatenation of a feature name and a corresponding feature value, and wherein the machine learning model implements a regularized logistic regression algorithm that includes ‘follow-the-regularized-leader’ (FTRL)-proximal learning that is trained based on associated hyperparameters to optimize learning accuracy on the set of enriched training feature vectors and corresponding interaction event tags; receiving, at one or more processors configured to execute the machine learning model, information relating to an online content placement slot and information relating to a user to whom content within the online content placement slot will be displayed within the defined time period; computing, by the one or more processors, an enriched estimation feature vector based upon: (i) a content item selected for placement within the online content placement slot that will be displayed within the defined time period, (ii) the information relating to the user, and (iii) the information relating to the online content placement slot; generating, for each feature value of the enriched estimation feature vector, a corresponding key; retrieving from the dictionary data structure, for each generated key, a corresponding coefficient value; determining, by the one or more processors executing the machine learning model, an estimate of likelihood of the user interacting with the selected content item within the defined time period, based upon the enriched estimation feature vector and the retrieved corresponding coefficient value, wherein the estimate of likelihood of the user interacting with the selected content item comprises a value representing a probability that the user will interact with the selected content item within the defined time period; and determining, based upon the estimate of likelihood of the user interacting with the selected content item within the defined time period, a bid amount for the selected content item for placement within the online content placement slot that will be displayed within the defined time period based on a magnitude of the value with respect to a bid threshold.
This invention relates to digital advertising and the use of machine learning to optimize content placement and bidding strategies. The system addresses the challenge of predicting user interaction with online content to improve ad performance and bidding efficiency. The method involves collecting and analyzing data from content placement events and user interaction events within a defined time window. These events are matched to generate a dataset where each record includes raw feature values from a content placement event and an interaction tag indicating whether a user interacted with the content. The raw features are transformed into enriched training feature vectors, which are used to train a logistic regression model. The model employs a regularized logistic regression algorithm with FTRL-proximal learning, optimized through hyperparameters to maximize accuracy. The trained model stores model coefficients in a dictionary, where each entry is keyed by a hashed combination of feature names and values. When a content placement opportunity arises, the system computes an enriched estimation feature vector based on the content, user data, and placement details. The model then generates a probability estimate of user interaction by retrieving corresponding coefficients from the dictionary. This probability is used to determine a bid amount for the content, ensuring competitive yet cost-effective bidding based on predicted engagement likelihood. The approach enhances ad targeting and bidding efficiency by leveraging machine learning to predict user behavior.
4. The method of claim 3 wherein: the online content placement slot is an ad slot; the information relating to the ad slot and information relating to the user to whom content within the ad slot will be displayed is received along with a bid request message transmitted from an ad exchange server; and the content item comprises at least one offer for placement within the ad slot.
5. The method of claim 4 further comprising: transmitting, to the ad exchange server by the one or more processors, a bid response message in reply to the bid request message; receiving, by the one or more processors from the ad exchange server, a successful bid notification; and updating, by the one or more processors, the online data store with content placement event data relating to placement of the content item.
This invention relates to digital advertising systems, specifically methods for processing bid requests and responses in real-time ad exchanges. The technology addresses inefficiencies in programmatic advertising where advertisers and publishers struggle to efficiently match ads with available inventory in real-time bidding environments. The system involves a processor that receives a bid request message from an ad exchange server, where the bid request includes details about an available ad placement opportunity, such as user data, ad space specifications, and pricing parameters. The processor evaluates the bid request against predefined criteria, such as targeting rules, budget constraints, and content relevance, to determine whether to submit a bid. If the criteria are met, the processor generates a bid response message and transmits it to the ad exchange server. Upon receiving a successful bid notification from the server, the processor updates an online data store with event data documenting the placement of the content item, including details like the winning bid amount, ad placement time, and user interaction metrics. This process ensures accurate tracking and reporting of ad placements in real-time bidding systems, improving transparency and efficiency in digital advertising transactions.
6. The method of claim 5 further comprising: receiving, by the one or more processors, a notification of a user interaction with the content item; and updating the online data store with user interaction event data relating to the user interaction with the content item.
7. The method of claim 6 wherein accessing the online data store, matching the retrieved content placement event records with the retrieved interaction event records, computing enriched training feature vectors, and training the machine learning model, are repeatedly executed to update the machine learning model.
8. A non-transitory computer readable storage medium comprising program code including instructions that, when executed by one or more processors of a demand side platform (DSP) server, cause the one or more processors of the DSP server to: implement a machine learning model which is executable by the one or more processors of the DSP server and configured to determine an estimate of likelihood of user interaction with a content item, the model having been trained using a set of enriched training feature vectors and corresponding interaction event tags derived from a matched data set generated from records relating to content placement events and records relating to user interaction events retrieved from an online data store, wherein the machine learning model is a logistic regression model comprising a plurality of model coefficients that are stored in a dictionary data structure in which each entry is defined by a key and a coefficient value, and each key comprises a hashed representation of a concatenation of a feature name and a corresponding feature value, wherein the machine learning model implements a regularized logistic regression algorithm that includes ‘follow-the-regularized-leader’ (FTRL)-proximal learning that is trained based on associated hyperparameters to optimize learning accuracy on the set of enriched training feature vectors and corresponding interaction event tags, wherein each record of the matched data set includes a set of raw feature values derived from an associated content placement event along with a corresponding interaction event tag indicating whether or not an interaction event occurred corresponding with the content placement event, wherein the content placement events and the user interaction events occur within a defined time period, and wherein the instructions, when executed by the one or more processors of the DSP server, cause the one or more processors of the DSP server to: receive, via the data communications interface, information relating to an online content placement slot; receive, via the data communications interface, information relating to a user to whom content within the online content placement slot will be displayed within the defined time period; compute an enriched estimation feature vector based upon: (i) a content item selected for placement within the online content placement slot that will be displayed within the defined time period, (ii) the information relating to the user, and (iii) the information relating to the online content placement slot; generate, for each feature value of the enriched estimation feature vector, a corresponding key; retrieve from the dictionary data structure, for each generated key, a corresponding coefficient value; determine, by the one or more processors of the DSP server executing the machine learning model, an estimate of likelihood of the user interacting with the selected content item within the defined time period based upon the enriched estimation feature vector and the retrieved corresponding coefficient value, wherein the estimate of likelihood of the user interacting with the selected content item comprises a value representing a probability that the user will interact with the selected content item within the defined time period; and determine, based upon the estimate of likelihood of the user interacting with the selected content item within the defined time period, a bid amount for the selected content item for placement within the online content placement slot that will be displayed within the defined time period based on a magnitude of the value with respect to a bid threshold.
Unknown
March 9, 2021
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.